Documentation

To get started with CopilotKit, please check out the documentation.

LangFuse Logging Integration

CopilotKit now supports LangFuse logging integration to help you monitor, analyze, and debug your LLM requests and responses.

Setup

To enable LangFuse logging, you can configure it when initializing the CopilotRuntime:

1import { CopilotRuntime, OpenAIAdapter } from "@copilotkit/runtime";
2import { LangfuseClient } from "langfuse";
3
4// Initialize your LangFuse client
5const langfuse = new LangfuseClient({
6  publicKey: process.env.LANGFUSE_PUBLIC_KEY!,
7  secretKey: process.env.LANGFUSE_SECRET_KEY!,
8  baseUrl: process.env.LANGFUSE_BASE_URL,
9});
10
11// Create a CopilotRuntime with LangFuse logging enabled
12const runtime = new CopilotRuntime({
13  adapter: new OpenAIAdapter({ apiKey: process.env.OPENAI_API_KEY }),
14  logging: {
15    enabled: true,
16    progressive: true, // Set to false for buffered logging
17    logger: {
18      logRequest: (data) => langfuse.trace({ name: "LLM Request", input: data }),
19      logResponse: (data) => langfuse.trace({ name: "LLM Response", output: data }),
20      logError: (errorData) => langfuse.trace({ name: "LLM Error", metadata: errorData }),
21    },
22  },
23});

Configuration Options

The logging configuration accepts the following options:

• enabled (boolean): Enable or disable logging (default: false)
• progressive (boolean): When true, logs each chunk as it's streamed. When false, logs the complete response (default: true)
• logger (object): Contains callback functions for logging:
  • logRequest: Called when an LLM request is made
  • logResponse: Called when an LLM response is received
  • logError: Called when an error occurs during an LLM request

Custom Logging Integrations

You can integrate with any logging service by implementing the logger interface:

1const runtime = new CopilotRuntime({
2  adapter: new OpenAIAdapter({ apiKey: "YOUR_API_KEY" }),
3  logging: {
4    enabled: true,
5    progressive: false,
6    logger: {
7      logRequest: (data) => {
8        // Implement your custom logging logic
9        console.log("LLM Request:", JSON.stringify(data));
10      },
11      logResponse: (data) => {
12        // Implement your custom logging logic
13        console.log("LLM Response:", JSON.stringify(data));
14      },
15      logError: (error) => {
16        // Implement your custom error logging
17        console.error("LLM Error:", error);
18      },
19    },
20  },
21});

This allows you to send your logs to any system or service that you prefer.